Forecasting of rainfall using different input selection methods on climate signals for neural network inputs

Long-term prediction of precipitation in planning and managing water resources, especially in arid and semi-arid countries such as Iran, has great importance. In this paper, a method for predicting long-term precipitation using weather signals and artificial neural networks is presented. For this purpose, climate data (large-scale signals) and meteorological data (local precipitation and temperature) with 3 to 12 months lead-times are used as inputs to predict precipitation for 3, 6, 9 and 12 months periods in 6 selected stations across Iran. A genetic algorithm (GA) and self-organized neural network (SOM) along with the application of winGamma software were comparatively used as input selection methods to choose the appropriate input variables. Examining the results, out of 96 predictions performed at all stations, in 43 cases, GA, in 28 cases, winGamma, and in 25 cases SOM have the best results compared to the other two methods. According to this, as a generalized assumption, it can be said that at least for the selected stations in this paper, the GA method is more reliable than the other two methods, and can be used to make predictions for future applications as a reliable input selection method. Moreover, among different climatic signals, Pacific Decadal Oscillation (PDO), Trans-Niño Index (TNI) and Eastern Tropical Pacific SST (NINO3) are the most repetitive indices for the most accurate forecast of each station.